Fast Outcome Categorization of the Upper Limb After Stroke.

Area of Science:

• Neurology
• Rehabilitation Medicine
• Biostatistics

Background:

• The Action Research Arm Test (ARAT) is a standard for assessing upper limb motor function after stroke.
• The COVID-19 pandemic highlighted the need for rapid, remote patient assessments.
• Current ARAT categorization involves 3, 4, or 5 outcome levels.

Purpose of the Study:

• To develop and validate decision trees using a subset of ARAT tasks.
• To create a faster method for categorizing upper limb motor outcomes post-stroke.
• To assess the accuracy of these new decision trees.

Main Methods:

• Retrospective analysis of ARAT scores from 333 stroke patients at 3 months poststroke.
• Utilized classification and regression tree analyses to identify key ARAT tasks.
• Validated the derived decision trees (FOCUS assessments) on a separate cohort of 157 patients at 6 months poststroke.

Main Results:

• Decision trees (FOCUS assessments) required only 2 to 4 ARAT tasks for categorization.
• Achieved high cross-validated accuracy, ranging from 87.7% to 96.7%.
• Maintained accuracy when applied to 6-month poststroke data (83.4%-91.7%).

Conclusions:

• A subset of ARAT tasks can effectively categorize upper limb motor outcomes after stroke.
• The developed FOCUS assessments offer an accurate and potentially faster alternative.
• Future research should explore remote application of FOCUS assessments via video call.